Interactive Clustering with a High-Performance ML Toolkit

Clustering is a class of machine learning algorithms which has important applications in many different fields. Users often use clustering to find hidden structures from data for those domain specific problems. However, evaluating clustering results is always a hard problem. In many and perhaps most of these applications, users need to trade off competing goals and encode prior knowledge into the model to define what is the best result. The learning algorithm however has evolved around the optimization of a single, usually narrowly-defined criterion, which may not obtain satisfactory results. In most cases, an expert makes trade-offs between different criteria which requires high-level (human) intelligence. This motivates us to provide interactive customization and optimization so that the expert can incorporate secondary criteria into the model-generation process in an interactive way. In this demo paper we will demonstrate the techniques we developed to do customized and interactive model optimization for clustering algorithms. The keys to the approach are (i) high-performance training so that non-trivial models can be trained in real-time (using roofline design and GPU hardware), (ii) a machine learning architecture which is modular, and supports primary and secondary loss functions, and (iii) highly-interactive visualization tools that support dynamic creation of visualizations and controls to match the bespoke criteria being optimized.